Adaptive segmentation of ultrasound images

摘要: Abstract This article describes a novel approach to the semi-automatic segmentation of ultrasound images. Assisted is particularly attractive when processing many slices through 3D data set, and even though fully automatic would be ideal, this currently not feasible given quality The algorithm developed in based on active contour paradigm, with several important modifications. attracted boundaries described locally by statistical models: allows for fact that definition what constitutes boundary may vary around boundary’s length. models are trained on-the-fly observing accepted operator. In way, operator intervention particular slice sensibly exploited reduce need subsequent slices. resulting provides fast, reliable verifiable vivo